No such column error lacks accessibility context for screen reader users

Direct Answers

1. Should ClickHouse implement enhanced error message formatting with accessibility metadata? Yes, but no such feature exists in any released version as of 2024. The HTTP and Native protocols return plain-text or JSON error objects containing only code, message, and stack_trace fields — no ARIA labels, remediation hints, or screen-reader annotations.
2. Would configurable error verbosity levels help? Partially. ClickHouse has send_logs_level and log_level settings, but these control logging volume, not error-message structure. They cannot add accessibility metadata to the error payload returned to the client.
3. How might the query parser be extended? The parser throws via Exception::create() with ErrorCodes::NO_SUCH_COLUMN_IN_TABLE (code 47). Error formatting is centralized in formatErrorMessage(). An extension would require modifying this function to accept an optional accessibility context object and propagating it through the protocol layer — a non-trivial change affecting all client interfaces.

Confirmed Facts

• Default identifier matching is case-sensitive; unquoted identifiers are normalized to lowercase. A column created as \"MixedCase\" must be referenced with double quotes or it resolves to mixedcase, triggering "No such column".
• The case_insensitive_identifiers=1 setting (added in 23.8) changes parser behavior globally but does not enrich error output.
• HTTP interface with default_format=JSON returns {\"code\":47,\"message\":\"DB::Exception: No such column: ColumnName\",...} — no accessibility fields.
• No ClickHouse GitHub issue or RFC proposes accessibility metadata for errors as of 2024.

Likely Explanation

ClickHouse's error system was designed for programmatic consumption (code + message) and operational logging. Screen-reader compatibility depends on the client application (clickhouse-client, JDBC driver, HTTP wrapper) announcing the error text. Server-side changes alone are insufficient — the client must expose the structured data to assistive technology. The project has prioritized parser correctness and performance over error-message richness.

Practical Workarounds Today

1. Enable query logging to capture failed statements for post-hoc analysis:

   SET log_queries = 1;

   Then query system.query_log for the failed statement and error.
2. Wrap client calls in application code to parse the error code (47) and re-emit a richer message:

   try { client.execute(query); }\ncatch (ClickHouseException e) {\n  if (e.code == 47) {\n    announce(\"Column not found. Check case sensitivity: \" +\n             \"unquoted identifiers are lowercased. \" +\n             \"Use double quotes for mixed-case names.\");\n  }\n}

3. Use the case-insensitive setting for development environments only:

   SET case_insensitive_identifiers = 1;

   This avoids the error class entirely but changes query semantics — not recommended for production without testing.

One Diagnostic Detail Needed

Which client interface are your screen-reader users employing? (clickhouse-client CLI, JDBC/ODBC driver, HTTP API wrapper, or a custom application?) The answer determines whether a client-side wrapper can inject accessibility context or whether a server-side protocol change is the only path.